Reinforcement learning for time optimal end-to-end quadcopter control
Main notebooks:
3D quad race:
- gym environment of end-to-end quadcopter model
- bebop quadcopter model from https://arxiv.org/pdf/2304.13460.pdf
- learned residual model (trained in the NNDroneModel notebook)
- constant disturbances
- training pipeline
- automatic c code generation
3D quad race INDI inner loop:
- gym environment of quadcopter model with thrust and rate inputs
- INDI controller is modeled as first order delay
- training pipeline
- automatic c code generation